Wow, it’s been over three months since I last posted anything. That reflects that I’ve been, mostly, working very damn hard and ending my days with little energy or time for a blog. However, this won’t do!
It’s been another long day and I’m tired and do need to sleep but I think this can be a short post and I think the reflection (in the title line of this post), was quite an interesting one, to me at least.
To begin at the beginning: I think I had two lectures on statistics as preclinical medical student and that’s the total of the formal teaching I’ve had on the subject. However, a huge proportion of the long hours I working at the moment are doing number crunching: statistics and psychometrics. Over the last 33 years in which I’ve been doing research, I did the number work on virtually all the papers I’ve co-authored and the majority of my papers, a not too shabby 132 peer-reviewed ones now, have been quantitative. Now that I’m no longer spending 40-60% of my week doing clinical work, I could be mistaken for a statistician, though I’ve always been very careful to say that I’m only an amateur at that game out of respect for anyone who is properly qualified and registered with one of the organisations like the UK’s Royal Statistical Society.
Earlier this week I was having a work zoom session with someone about a possible collaboration. I’ve become very wary about what I agree to do these days and will only do the statistics for a piece of work if I have pretty total control over how that bit of the work is done and a full say in how the findings are interpreted in the discussion part of the paper. It’s that last bit that’s got me the revelation about my identity. It’s also that bit of my insistence on how I work with people that is sometimes tricky, as I tend to be very against any overstating of things or any minimising of the caveats and concerns.
In the conversation I heard myself say “that’s because I’m really a researcher not a statistician” and it echoed in my head as having been more true and more important than I’d realised as I said it.
Statisticians don’t like it, but too often they don’t get a say in the design of a paper. However, in my limited experience of working with professional statisticians, I think they’re often quite happy not having much or anything to do with the discussion, except perhaps an initial translation of the findings into words rather than graphs and tables. Even that translation is often kept entirely in the results section of a paper.
What I realised was that I really want to be involved in the complete sequence that should be the skeleton of any quantitative paper:
- Agreed, explicit aims/objectives that drive the …
- … sampling design and data collection which is is part of, and congruent with …
- … the plan of analysis, defined “a priori“, i.e. before seeing the data and including a “stopping rule” also defined a priori so you can show you didn’t keep looking at your accumulating data until it, perhaps by chance, said what you wanted it to say, or just seemed to say something interesting (not necessarily a fetishised “statistical power calculation” but a definite and sensible stopping rule) and then ..
- … the actual analyses, saying clearly when some analysis wasn’t part of the a priori plan but was a sensible pursuit of more clarity around something emergent in the data that you hadn’t expected, so much so that you hadn’t planned for it in your plan of analyses, all this leading into …
- … a discussion, with caveats, perhaps some “conclusions” and perhaps some implications for the subject area, in the case of my research, for clinical practice or the ways we research it.
I really do love the mathematical bit in items 3 and 4 of that sequence, however, what really motivates me is know I had a part in, and a responsibility for, trying to make the whole sequence as honest and as useful as possible.
My life would be simpler were I happy to confine myself to items 3 and 4 … but I’m not.
Some other night. I think this might lead me to write about “American football numbers” (the ones on the players backs, not the incredible plethora of numbers that the game uses and sometimes calls “statistics”). I think I should also link this “researcher not statistician” issue with the similar “clinical researcher not researcher” issue; that, for me, isn’t just about my topic area or focus. Oh dear, I can see that leads to one theme I’ve being processing for over two years now in this very erratic blog: how the “clinically retired” collides with the “clinical researcher”. But enough for now, as I think I often finish up here: “To sleep, perchance to dream!” Oh aye there is a rub but also lovely link back to one of the great statistical quote of all times:
To consult the statistician after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment died of.
Ronald Fisher. Presidential Address to the First Indian Statistical Congress, 1938. Sankhya 4, 14-17. [https://en.wikiquote.org/wiki/Ronald_Fisher].
[Added 6.xi.18. This was a bit dry so here’s an icicle for visual amusement; and I have followed on to What does it mean to be a “clinical researcher” not “researcher” .]
One thought on “Realising that I’m a (retired clinical) researcher, not an amateur statistician”